Generation of a computer program to test for correct operation of a data processing apparatus

Abstract

Software built in self test computer programs 12 are generated using a genetic algorithm 14 technique. A fault simulator 20 is used to simulate candidate software built in self test computer programs and compare the simulated execution, such to deliberately introduced test faults, with expected execution outcomes previously derived for that candidate program to determine the sensitivity of that candidate program to the faults which are introduced. This score can be fed back into the genetic algorithm mutation to converge the mutation process upon appropriately fault sensitive software built in self test program code.

Description

BACKGROUND OF THE INVENTION [0001] 1. Field of the Invention [0002] This invention relates to the field of data processing systems. More particularly, this invention relates to the generation of computer programs for testing data processing apparatus for correct operation. [0003] 2. Description of the Prior Art [0004] The testing of data processing apparatus for correct operation is an important area. Data processing systems are increasingly being used in critical environments where it is desirable to be able to regularly check that those data processing systems are operating correctly and continue to be fault free. With the this in mind, it is known to provide data processing systems with built in self test (BIST) capabilities such that, for example, when the data processing system is initialised it performs hardware tests upon itself to check for correct operation. It may additionally be advisable to periodically perform tests upon the data processing apparatus to ensure that it c...

AI Technical Summary

Benefits of technology

[0015] The present technique recognises that genetic algorithms can be employed to automatically generate computer programs for testing a data processing apparatus for correct operation with the ability of those computer programs to be sensitive to faults simulated upon the data processing apparatus being made a convergence criteria in a way that allows the technique to produce effective test computer programs in a reasonable amount of time. Furthermore, the test computer programs generated in this way can have an objective assessment of their sensitivity to fault conditions. This is particularly useful in safety critical systems, such as automotive braking systems, where objective information concerning fault management is highly beneficial.

[0016] The process of generating the candidate computer programs with their associated expected execution outcome can be speeded up by using an instruction set simulator to simulate their execution and generate the expected execution outcome data.

[0018] The formation of the candidate computer programs and the associated execution outcome can also be conducted using an RTL (register transfer language) simulator. Whilst such a simulator will typically be slower than an instruction set simulator, and accordingly desirable to use after the instruction set simulator has produced suitable good candidates, an RTL simulator has the advantage of being able to produce an additional type of scoring based upon a breadth of RTL line coverage or RTL register value toggling measured during simulation of the candidate computer programs with the RTL simulator.

[0020] The generation technique may be automated in a way that reduces the overhead associated with the production of such test computer programs and the system can be allowed to iterate until appropriate termination conditions are reached, such as a probability of producing a detectable change in execution outcome exceeding a predetermined level, such a probability ceasing to improve at a rate above a certain rate (indicative of convergence upon a solution), the number of iterations being performed exceeding a predetermined number or the time spent iterating the solutions exceeding a predetermined time.

Problems solved by technology

As the complexity of data processing systems increases, the difficulty of writing software to meet the above criteria also increases and is a task which can take a highly skilled software engineer a disadvantageously long time to complete.

Such a software engineer can use their skill and knowledge of the data processing system concerned to exercise a wide variety of its functional capabilities with such hand written code, but even when this considerable effort is undertaken it can be difficult to verify objectively how thorough such test code is in exercising the data processing system and being sensitive to possible faults.

Images